CN110633751A - Training method of car logo classification model, car logo identification method, device and equipment - Google Patents
Training method of car logo classification model, car logo identification method, device and equipment Download PDFInfo
- Publication number
- CN110633751A CN110633751A CN201910875817.6A CN201910875817A CN110633751A CN 110633751 A CN110633751 A CN 110633751A CN 201910875817 A CN201910875817 A CN 201910875817A CN 110633751 A CN110633751 A CN 110633751A
- Authority
- CN
- China
- Prior art keywords
- classification
- fusion
- loss
- vehicle
- logo
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a training method of a car logo classification model, a car logo identification method, a device and equipment. The training method of the car logo classification model comprises the following steps: acquiring a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; inputting the fused sample image into a car logo classification model for classification processing to obtain a car logo classification result; respectively calculating a first classification loss and a second classification loss of the vehicle logo classification result and the class labels of any two training sample images, and fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss; and adjusting model parameters in the car logo classification model according to the fusion loss, and performing circular training until the car logo classification model converges. The method can realize continuous discrete samples, improve the smoothness in the neighborhood and solve the problem of overfitting; and simultaneously improves the efficiency of model training.
Description
Technical Field
The application relates to the technical field of image recognition, in particular to a training method of a car logo classification model, a car logo recognition method, a device and equipment.
Background
Along with the increasing living standard of people, the number of motor vehicles is increased sharply, the workload of vehicle annual inspection of a vehicle management station is increased, and vehicle identification is a more important part in the vehicle annual inspection process. Early car logo identification work is usually judged by manpower, but the labor cost is high, the influence of subjective factors is large, and the efficiency is low. At present, the field of artificial intelligence is widely applied, and the identification efficiency and accuracy can be greatly improved by utilizing an artificial intelligence mode (such as using a neural network model) to identify the car logo.
Before the car logo is identified by using the neural network model, the neural network model needs to be trained, and the traditional technology generally includes that car logo labeling is carried out on archive pictures of a vehicle, and then the pictures and corresponding labels are input into the neural network model for training. However, the model training method in the conventional technology has the over-fitting problem, and when the car logo area of the car picture acquired in the annual inspection process of the car is not clear, the accuracy of car logo recognition by using the trained neural network model is low.
Disclosure of Invention
Therefore, it is necessary to provide a training method, a car logo recognition method, a device and equipment for a car logo classification model, for solving the problem of low accuracy of car logo recognition by using a trained neural network model in the conventional technology.
In a first aspect, an embodiment of the present application provides a training method for a car logo classification model, including:
acquiring a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; the training sample image comprises a category label;
inputting the fused sample image into a car logo classification model for classification processing to obtain a car logo classification result;
respectively calculating a first classification loss and a second classification loss of the vehicle logo classification result and the class labels of any two training sample images, and fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss;
and adjusting model parameters in the car logo classification model according to the fusion loss, and performing circular training until the car logo classification model converges.
In one embodiment, fusing any two training sample images in the plurality of training sample images according to a preset fusion ratio to obtain a fused sample image, including:
multiplying a first training sample image in any two training sample images by the fusion proportion to obtain a first result;
multiplying a second training sample image in any two training sample images by the deformation form of the fusion proportion to obtain a second result;
and adding the first result and the second result to obtain a fused sample image.
In one embodiment, the car logo classification result is a probability vector of the fused sample image belonging to different car logo classes; respectively calculating the first classification loss and the second classification loss of the car logo classification result and the class labels of any two training sample images, and the method comprises the following steps:
respectively carrying out one-hot transformation on the class labels of any two training sample images to obtain a first class vector and a second class vector;
and calculating a first classification loss of the vehicle logo classification result and the first class vector and a second classification loss of the vehicle logo classification result and the second class vector.
In one embodiment, fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss, including:
multiplying the first classification loss by the fusion proportion to obtain a third result; multiplying the second classification loss by the deformation form of the fusion proportion to obtain a fourth result;
and adding the third result and the fourth result to obtain fusion loss.
In one embodiment, fusing any two training sample images in the plurality of training sample images according to a preset fusion ratio to obtain a fused sample image, including:
according to a formula containing λ xi+(1-λ)xjObtaining a fused sample image according to the relation; wherein, λ is fusion ratio, xiFor the first training sample image, xjIs the second training sample image.
In one embodiment, fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss, including:
according to the formula containing λ cross EntroLoss (y', y)i)+(1-λ)*crossEntroLoss(y',yj) Obtaining the fusion loss according to the relation; wherein y' is the classification result of the car logo, yiIs a first class vector, yjAs a second class vector, crossEntroLoss (y', y)i) For the first classification loss, crossEntroLoss (y', y)j) Is the second classification loss.
In a second aspect, an embodiment of the present application provides a car logo identification method, including:
acquiring an image to be detected;
detecting an image to be detected by adopting a vehicle detection model, and marking a region of a vehicle if the vehicle is detected; if the vehicle is not detected, the process is ended;
detecting the marked vehicle area by adopting a vehicle logo detection model, and marking the area of the vehicle logo if the vehicle logo is detected; if the car logo is not detected, ending the process;
adopting a vehicle logo classification model to identify the marked vehicle logo area to obtain a vehicle logo identification result; the training method of the car logo classification model is the method in the embodiment.
In a third aspect, an embodiment of the present application provides a training apparatus for a car logo classification model, including:
the first acquisition module is used for acquiring a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; the training sample image comprises a category label;
the classification module is used for inputting the fusion sample image into the car logo classification model for classification processing to obtain a car logo classification result;
the calculation module is used for calculating the first classification loss and the second classification loss of the vehicle logo classification result and the class labels of any two training sample images respectively, and fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss;
and the training module is used for adjusting the model parameters in the vehicle logo classification model according to the fusion loss so as to carry out circular training until the vehicle logo classification model converges.
In a fourth aspect, an embodiment of the present application provides a car logo recognition device, including:
the second acquisition module is used for acquiring an image to be detected;
the first detection module is used for detecting the image to be detected by adopting a vehicle detection model, and marking the area of the vehicle if the vehicle is detected; if the vehicle is not detected, the process is ended;
the second detection module is used for detecting the marked vehicle area by adopting the vehicle logo detection model, and marking the marked vehicle area if the vehicle logo is detected; if the car logo is not detected, ending the process;
and the identification module is used for identifying the marked car logo area by adopting the car logo classification model to obtain a car logo identification result.
In a fifth aspect, an embodiment of the present application provides a computer device, including a memory and a processor, where the memory stores a computer program, and the processor implements the following steps when executing the computer program:
acquiring a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; the training sample image comprises a category label;
inputting the fused sample image into a car logo classification model for classification processing to obtain a car logo classification result;
respectively calculating a first classification loss and a second classification loss of the vehicle logo classification result and the class labels of any two training sample images, and fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss;
and adjusting model parameters in the car logo classification model according to the fusion loss, and performing circular training until the car logo classification model converges.
In a sixth aspect, an embodiment of the present application further provides a computer device, including a memory and a processor, where the memory stores a computer program, and the processor implements the following steps when executing the computer program:
acquiring an image to be detected;
detecting an image to be detected by adopting a vehicle detection model, and marking a region of a vehicle if the vehicle is detected; if the vehicle is not detected, the process is ended;
detecting the marked vehicle area by adopting a vehicle logo detection model, and marking the area of the vehicle logo if the vehicle logo is detected; if the car logo is not detected, ending the process;
and identifying the marked car logo area by adopting a car logo classification model to obtain a car logo identification result.
In a seventh aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the following steps:
acquiring a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; the training sample image comprises a category label;
inputting the fused sample image into a car logo classification model for classification processing to obtain a car logo classification result;
respectively calculating a first classification loss and a second classification loss of the vehicle logo classification result and the class labels of any two training sample images, and fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss;
and adjusting model parameters in the car logo classification model according to the fusion loss, and performing circular training until the car logo classification model converges.
In an eighth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the following steps:
acquiring an image to be detected;
detecting an image to be detected by adopting a vehicle detection model, and marking a region of a vehicle if the vehicle is detected; if the vehicle is not detected, the process is ended;
detecting the marked vehicle area by adopting a vehicle logo detection model, and marking the area of the vehicle logo if the vehicle logo is detected; if the car logo is not detected, ending the process;
and identifying the marked car logo area by adopting a car logo classification model to obtain a car logo identification result.
Firstly, obtaining a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion ratio to obtain a fusion sample image; the training sample image comprises a category label; then inputting the fused sample image into a car logo classification model for classification processing to obtain a car logo classification result; respectively calculating a first classification loss and a second classification loss of the vehicle logo classification result and the class labels of any two training sample images, and fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss; and finally, adjusting model parameters in the vehicle logo classification model according to the fusion loss, and performing circular training until the vehicle logo classification model converges. According to the method, any two images are fused according to a fusion ratio, so that discrete samples can be serialized, the smoothness in the neighborhood is improved, and the problem of overfitting is solved; and data can be fused before model training, so that a preprocessing mode in model training is simplified, and the efficiency of model training is improved.
Drawings
FIG. 1 is a schematic flowchart illustrating a training method of a car logo classification model according to an embodiment;
FIG. 2 is a schematic flowchart of a training method for a car logo classification model according to another embodiment;
FIG. 2a is an exemplary diagram of a fused sample image obtained by fusing two training sample images according to an embodiment;
FIG. 3 is a schematic flowchart of a training method for a car logo classification model according to another embodiment;
FIG. 4 is a schematic flow chart illustrating a vehicle logo identification method according to an embodiment;
FIG. 5 is a schematic structural diagram of a training apparatus for a car logo classification model according to an embodiment;
FIG. 6 is a schematic structural diagram of an emblem identification device according to an embodiment;
fig. 7 is a schematic internal structural diagram of a computer device according to an embodiment.
Detailed Description
The training method for the car logo classification model provided by the embodiment of the application can be suitable for the training process of the classification network model, the classification network model can be used for identifying and classifying car logo images, and can also be used for classifying other images, such as the classification of sportsman ball cover numbers and the like, and the embodiment does not limit the method. In the traditional technology, when a car logo classification model is trained, a car archive picture or a field acquired image is usually used for training, but the problem of overfitting exists in the training method due to the fact that the car image is clear, and if the trained car logo classification model is used for identifying an unclear car logo image, the identification accuracy is low. The embodiment of the application provides a training method of a car logo classification model, a car logo identification method, a device and equipment, and aims to solve the technical problems.
In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions in the embodiments of the present application are further described in detail by the following embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
It should be noted that the execution subject of the method embodiments described below may be a training apparatus for the car logo classification model, and the apparatus may be implemented as part of or all of a computer device by software, hardware, or a combination of software and hardware. The following method embodiments are described by taking the execution subject as the computer device as an example.
Fig. 1 is a schematic flowchart of a training method of a car logo classification model according to an embodiment. The embodiment relates to a specific process of training a car logo classification model by utilizing a training sample image through computer equipment. As shown in fig. 1, the method includes:
s101, obtaining a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; the training sample image includes a class label.
Specifically, the training sample images are a large number of car logo area images of different types, different colors, different shooting angles and different illuminations, and each training sample image is marked with a category label. Assume a total of n training sample images, m class labels, xiFor the ith training sample image (where i ≦ n), yiIs xiCorresponding category labels (e.g. Volkswagen, Benz, Chery, etc., m in total), xjIs the jth training sample image (wherein j is less than or equal to n and i is not equal to j), yjIs xjCorresponding class label, the computer device can associate xiAnd xj(i.e. x)iAnd xjAny two different training sample images) are fused according to a preset fusion ratio lambda to obtain a fusion sample image. Alternatively, the data format of the category label may be a category label (e.g., a number), and different category labels represent different category labels. Optionally, the computer device may adopt a mixup data enhancement method to fuse any two training sample images.
Wherein x isiAnd xjThe images can be regarded as discrete training sample images, and after the discrete training sample images are fused according to a preset fusion ratio, a series of continuous training sample images can be obtained, so that the discrete samples can be continuous, and the smoothness in the neighborhood is improved.
Optionally, λ may be a preset value, or may be a random value between (0 and 1) calculated by using a beta (α, α) distribution (as calculated by experiments, α is 0.2, which is the best training effect), and of course, the value of λ may also be obtained by using other distribution calculation methods.
And S102, inputting the fused sample image into the car logo classification model for classification processing to obtain a car logo classification result.
Specifically, after obtaining the fusion sample image, the computer device may input the fusion sample image into a car logo classification model for classification processing, so as to obtain a car logo classification result, where the classification result is one of the m kinds of category labels. Optionally, the vehicle logo classification result obtained by the computer device may be probabilities that the fusion sample image belongs to different vehicle logo categories, and a sum of the probabilities is 1, for example, the probability of belonging to the public is 0.6, the probability of belonging to the galloping is 0.2, the probability of belonging to the curiosity is 0,. and the probability of belonging to the bmw is 0.2, and then the category corresponding to the maximum probability value is used as the vehicle logo classification result.
Optionally, the car logo classification model may be a Neural network model, such as a Convolutional Neural Network (CNN) and a Full Convolutional Network (FCN), or may be other deep learning models and machine learning models, which is not limited in this embodiment. Optionally, the vehicle logo classification model may use resnet-18 as a base network, where two full connection layers are connected behind the base network, and the size of the last full connection layer is equal to the number (i.e., m) of vehicle logo class labels.
S103, respectively calculating the first classification loss and the second classification loss of the car logo classification result and the class labels of any two training sample images, and fusing the first classification loss and the second classification loss according to a fusion ratio to obtain fusion loss.
Specifically, after obtaining the car logo classification result, the computer device may calculate a first classification loss and a second classification loss of the car logo classification result and the class labels of any two training sample images, that is, calculate the car logo classification result and yiThe first classification loss of (2), and the car logo classification result and yjIs lost. Then, since the fusion sample image is input into the car logo classification model, the computer device can also fuse the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss. Optionally, the first classification loss and the second classification loss are fusedThe fusion ratio of (b) may be λ, or may be another relational expression including λ.
Optionally, the computer device may use a cross entropy loss function as a loss function for calculating the first classification loss and the second classification loss, and of course, may also use other types of loss functions for calculation, which is not limited in this embodiment.
And S104, adjusting model parameters in the vehicle logo classification model according to the fusion loss, and performing circular training until the vehicle logo classification model converges.
Specifically, the computer device may adjust the model parameters in the vehicle logo classification model according to the fusion loss, and optionally, may adjust the model parameters in a reverse gradient propagation manner. And optionally, the condition of convergence of the vehicle logo classification model can be that the fusion loss reaches a preset loss threshold value, and can also be that the fusion loss reaches convergence.
Optionally, the computer device may set an initial learning rate in the car logo classification model training process to be 0.01, the learning rate is reduced along with the increase of the number of iterations in a multi-step manner, and the optimizer for adjusting the model parameters may be an SGD, Adam, or other type of optimizer, which is not limited in this embodiment.
In the training method for the car logo classification model provided by this embodiment, computer equipment firstly fuses any two images in a plurality of acquired training sample images according to a fusion proportion, then inputs the fusion sample images into the car logo classification model to obtain a car logo classification result, then respectively calculates a first classification loss and a second classification loss of the car logo classification result and class labels of any two training sample images, fuses the two losses, and finally adjusts model parameters in the car logo classification model according to the fusion loss, so that the cycle training is performed until the car logo classification model converges. According to the method, any two images are fused according to a fusion ratio, so that discrete samples can be serialized, the smoothness in the neighborhood is improved, and the problem of overfitting is solved; and data can be fused before model training, so that a preprocessing mode in model training is simplified, and the efficiency of model training is improved.
Optionally, in some embodiments, when fusing any two training sample images, the computer device may also correspondingly fuse the category labels corresponding to the two training sample images; and then calculating the loss between the car logo classification result and the label before fusion, and adjusting the model parameters in the car logo classification model by using the loss.
Fig. 2 is a schematic flowchart of a training method of the car logo classification model according to another embodiment. The embodiment relates to a specific process of fusing any two training sample images in a plurality of training sample images according to a preset fusion proportion by computer equipment to obtain a fused sample image. On the basis of the foregoing embodiment, optionally, S101 may include:
s201, multiplying a first training sample image in any two training sample images by the fusion proportion to obtain a first result.
S202, multiplying a second training sample image in any two training sample images by the transformation form of the fusion proportion to obtain a second result.
Specifically, after the computer device determines the fusion ratio λ, a first training sample image (i.e., one of the training sample images) in any two of the training sample images may be multiplied by the fusion ratio to obtain a first result; and multiplying the second training sample image (namely the other training sample image) by the deformation form of the fusion proportion to obtain a second result. Alternatively, the modified form of the fusion ratio λ may be (1- λ), or may be a weight of the fusion ratio.
Wherein the value of λ can be calculated by the above-mentioned beta (α, α) distribution, and optionally, the formula of the beta distribution can be
α ═ β, then the equation can yield oneThe λ is calculated by uniform distribution.
And S203, adding the first result and the second result to obtain a fused sample image.
Specifically, the computer device may add the first result and the second result to obtain a fused sample image, and optionally, may sum the first result and the second result directly, or may sum the first result and the second result in a weighted manner to obtain the fused sample image.
Alternatively, the computer device may be based on a computer system containing λ xi+(1-λ)xjObtaining a fused sample image, wherein lambda is a fusion proportion and x isiFor the first training sample image, λ xiAs a first result, xjFor the second training sample image, (1- λ) xjIs the second result. Optionally, the computer device may also include λ yi+(1-λ)yjObtaining the fusion label. An example of fusing two training sample images to obtain a fused sample image is shown in fig. 2a, and it should be noted that the car logo image is not shown in the figure as an example, which is only one example.
In the training method of the car logo classification model provided by the embodiment, the computer device multiplies the first training sample image by the fusion proportion to obtain a first result; multiplying the second training sample image by the deformation form of the fusion proportion to obtain a second result; and then adding the first result and the second result to obtain a fused sample image. The obtained fusion sample image enables the continuity of the training sample to be better, and the overfitting problem can be better solved.
Optionally, the car logo classification result may be a probability vector that the fusion sample image belongs to different car logo classes, and therefore if a first classification loss and a second classification loss of the car logo classification result and the class labels of any two training sample images are to be calculated, the class labels of the training sample images also need to be in a vector form. Optionally, as shown in fig. 3, a schematic flow chart of a training method of a car logo classification model according to another embodiment is provided. The embodiment relates to a specific process that a computer device respectively calculates a car logo classification result and a first classification loss and a second classification loss of class labels of any two training sample images. Optionally, the step S103 may include:
s301, performing one-hot transformation on the class labels of any two training sample images to obtain a first class vector and a second class vector.
S302, calculating a first classification loss of the vehicle logo classification result and the first category vector and a second classification loss of the vehicle logo classification result and the second category vector.
Specifically, the computer device may perform one-hot transformation on the category labels of any two training sample images, and if 5 car logo categories are assumed, the category label of the first training sample image is 2, and the category label of the first training sample image is 4, the first category vector of the first training sample image is [0, 1, 0, 0, 0], and the second category vector of the second training sample image is [0, 0, 0, 1, 0 ]. The computer device may then calculate a first classification loss of the car logo classification result and the first class vector, and a second classification loss of the car logo classification result and the second class vector, where the method for calculating the loss may be as described in the above embodiments.
Optionally, after the computer device calculates the first classification loss and the second classification loss, the first classification loss may be multiplied by the fusion proportion to obtain a third result, and the second classification loss may be multiplied by the deformation form of the fusion proportion to obtain a fourth result; and then adding the third result and the fourth result to obtain fusion loss. Alternatively, the computer device may be based on a system comprising λ × cross entroloss (y', y)i)+(1-λ)*crossEntroLoss(y',yj) Obtaining the fusion loss, wherein y' is the classification result of the vehicle logo predicted by the network, and yiIs a first class vector, yjAs a second class vector, crossEntroLoss (y', y)i) For the first classification loss, crossEntroLoss (y', y)j) Is the second classification loss.
In the training method for the car logo classification model provided in this embodiment, after the computer device performs one-hot transformation on the class labels of any two training sample images to obtain the first class vector and the second class vector, the computer device calculates the first classification loss of the car logo classification result and the first class vector and the second classification loss of the car logo classification result and the second class vector. Therefore, when the classification loss is calculated, the fact that the vehicle logo classification result and the category label have the same data format is guaranteed, the accuracy of fusion loss obtained through calculation is improved, and the precision of a vehicle logo classification model obtained through training is further improved.
The above embodiment describes a training process of the car logo classification model, and after the training is converged, the training process can be applied to an actual scene, such as a vehicle annual inspection process. Fig. 4 is a schematic flowchart of a car logo identification method according to an embodiment, and as shown in fig. 4, the method includes:
s401, acquiring an image to be detected.
Specifically, the computer device first obtains an image to be detected, optionally, the image to be detected may be an image taken at the vehicle annual inspection site, and the computer device may obtain the image to be detected at a fixed time interval from a memory storing the image to be detected.
S402a, detecting the image to be detected by adopting the vehicle detection model, and marking the area of the vehicle if the vehicle is detected.
S402b, if no vehicle is detected, the flow ends.
Specifically, the computer device may detect the image to be detected by using a vehicle detection model, and determine whether there is a vehicle region. The vehicle detection model may be a neural network model, may also be other deep learning models, and optionally may be an ssd (single Shot multi box detector) network model. If a vehicle region is detected in the image to be detected, the computer device marks the vehicle region, optionally, a rectangular frame mark may be adopted, a circular frame mark may be adopted, and certainly, other methods may also be adopted for marking. If the vehicle area is not detected in the image to be detected, the process is ended, and the computer equipment detects the next image to be detected.
Optionally, the training process of the vehicle detection model may be: obtaining vehicle images and archival images of a vehicle inspection site at different angles and under different illumination conditions, and marking the vehicles in the images by using rectangular frames (or other marking frames) as real labels; the labeled images are then used to train the vehicle detection model until the vehicle detection model converges. The initial learning rate can be set to be 0.01, the learning rate is reduced along with the increase of the iteration times in a multi-step mode, and the optimizer adopts SGD.
And S403a, detecting the marked vehicle area by using the vehicle logo detection model, and marking the vehicle logo area if the vehicle logo is detected.
S403b, if the emblem is not detected, the process ends.
Specifically, after the computer device obtains the marked vehicle region, the computer device may detect the vehicle region image by using the vehicle logo detection model, and determine whether there is a vehicle logo region. The car logo detection model may be a neural network model, may also be other deep learning models, and optionally may be an ssd (single Shot multi box detector) network model. If a vehicle logo area is detected in the vehicle area image, the computer equipment marks the vehicle logo area, optionally, a rectangular frame mark or a round frame mark can be adopted, and certainly, other modes can be adopted for marking; optionally, the marking mode of the car logo area may be the same as or different from that of the vehicle area. If the vehicle logo area is not detected in the vehicle area image, the process is ended, and the computer equipment detects the next image to be detected.
Optionally, the training process of the car logo detection model may be as follows: obtaining vehicle images and archive images of a vehicle inspection site at different angles and under different illumination conditions, and marking a vehicle logo in the images by using a rectangular frame (or other marking frames) as a real label; and then training the car logo detection model by using the marked images until the car logo detection model converges. The initial learning rate can be set to be 0.01, the learning rate is reduced along with the increase of the iteration times in a multi-step mode, and the optimizer adopts SGD.
S404, identifying the marked car logo area by adopting a car logo classification model to obtain a car logo identification result; the training method of the car logo classification model is the method described in the above embodiment.
Specifically, the computer device may identify the marked car logo region by using the trained car logo classification model to obtain a car logo identification result. Optionally, after the vehicle logo identification result is obtained, the computer device may match the vehicle logo identification result with the information registered in the archive, mark the vehicle logo as 1 if the matching is successful, and mark the vehicle logo as 0 if the matching is unsuccessful, so as to obtain an annual inspection result of the vehicle.
In the vehicle logo identification method provided by the embodiment, the computer equipment firstly detects the vehicle of the image to be detected and marks the area of the vehicle; then, carrying out vehicle logo detection on the marked vehicle area, and marking the area of the vehicle logo; and finally, carrying out vehicle logo identification on the marked vehicle logo area. Therefore, the accuracy of vehicle logo identification can be improved by layer-by-layer detection and identification.
It should be understood that although the various steps in the flowcharts of fig. 1-4 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 1-4 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternating with other steps or at least some of the sub-steps or stages of other steps.
Fig. 5 is a schematic structural diagram of a training device for a car logo classification model according to an embodiment. As shown in fig. 5, the apparatus includes: a first acquisition module 11, a classification module 12, a calculation module 13 and a training module 14.
Specifically, the first obtaining module 11 is configured to obtain a plurality of training sample images, and fuse any two training sample images in the plurality of training sample images according to a preset fusion ratio to obtain a fusion sample image; the training sample image includes a class label.
And the classification module 12 is used for inputting the fused sample image into the car logo classification model for classification processing to obtain a car logo classification result.
And the calculating module 13 is configured to calculate a first classification loss and a second classification loss of the car logo classification result and the class labels of any two training sample images, and fuse the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss.
And the training module 14 is used for adjusting model parameters in the vehicle logo classification model according to the fusion loss, and performing circular training until the vehicle logo classification model converges.
The training device for the car logo classification model provided by the embodiment can execute the method embodiment, and the implementation principle and the technical effect are similar, and are not repeated herein.
In one embodiment, the first obtaining module 11 is specifically configured to multiply a first training sample image of any two training sample images by a fusion ratio to obtain a first result; multiplying a second training sample image in any two training sample images by the deformation form of the fusion proportion to obtain a second result; and adding the first result and the second result to obtain a fused sample image.
In one embodiment, the car logo classification result is a probability vector of the fused sample image belonging to different car logo classes; the calculating module 13 is specifically configured to perform one-hot transformation on the category labels of any two training sample images, so as to obtain a first category vector and a second category vector; and calculating a first classification loss of the vehicle logo classification result and the first category vector and a second classification loss of the vehicle logo classification result and the second category vector.
In one embodiment, the calculating module 13 is specifically configured to multiply the first classification loss by the fusion ratio to obtain a third result; multiplying the second classification loss by the deformation form of the fusion proportion to obtain a fourth result; and adding the third result and the fourth result to obtain the fusion loss.
In one embodiment, the first obtaining module 11 is specifically configured to obtain the value according to a rule including λ xi+(1-λ)xjObtaining a fused sample image according to the relation; wherein, λ is fusion ratio, xiFor the first training sample image, xjIs the second training sample image.
In one embodiment, the calculating module 13 is specifically configured to calculate the value according to a method including λ × cross entroloss (y', y)i)+(1-λ)*crossEntroLoss(y',yj) Obtaining the fusion loss according to the relation; wherein y' is the classification result of the car logo, yiIs a first class vector, yjAs a second class vector, crossEntroLoss (y', y)i) For the first classification loss, crossEntroLoss (y', y)j) Is the second classification loss.
Fig. 6 is a schematic structural diagram of a car logo identification device according to an embodiment. As shown in fig. 6, the apparatus includes: a second acquisition module 21, a first detection module 22, a second detection module 23 and an identification module 24.
Specifically, the second obtaining module 21 is configured to obtain an image to be detected.
The first detection module 22 is configured to detect an image to be detected by using a vehicle detection model, and mark a region of a vehicle if the vehicle is detected; if the vehicle is not detected, the flow ends.
The second detection module 23 is configured to detect the marked vehicle area by using the vehicle logo detection model, and mark the marked vehicle area if the vehicle logo is detected; if the car logo is not detected, the process is ended.
And the identification module 24 is used for identifying the marked car logo area by adopting the car logo classification model to obtain a car logo identification result.
The car logo recognition device provided by the embodiment can execute the method embodiment, and the implementation principle and the technical effect are similar, so that the details are not repeated.
For the specific limitations of the training device and the car logo recognition device of the car logo classification model, reference may be made to the above limitations of the training method and the car logo recognition method of the car logo classification model, and details are not repeated here. The modules in the training device and the car logo recognition device of the car logo classification model can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 7. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a training method or a car logo recognition method for a car logo classification model. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 7 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:
acquiring a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; the training sample image comprises a category label;
inputting the fused sample image into a car logo classification model for classification processing to obtain a car logo classification result;
respectively calculating a first classification loss and a second classification loss of the vehicle logo classification result and the class labels of any two training sample images, and fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss;
and adjusting model parameters in the car logo classification model according to the fusion loss, and performing circular training until the car logo classification model converges.
The implementation principle and technical effect of the computer device provided in this embodiment are similar to those of the method embodiments described above, and are not described herein again.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
multiplying a first training sample image in any two training sample images by the fusion proportion to obtain a first result;
multiplying a second training sample image in any two training sample images by the deformation form of the fusion proportion to obtain a second result;
and adding the first result and the second result to obtain a fused sample image.
In one embodiment, the car logo classification result is a probability vector of the fused sample image belonging to different car logo classes; the processor, when executing the computer program, further performs the steps of:
respectively carrying out one-hot transformation on the class labels of any two training sample images to obtain a first class vector and a second class vector;
and calculating a first classification loss of the vehicle logo classification result and the first class vector and a second classification loss of the vehicle logo classification result and the second class vector.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
multiplying the first classification loss by the fusion proportion to obtain a third result; multiplying the second classification loss by the deformation form of the fusion proportion to obtain a fourth result;
and adding the third result and the fourth result to obtain fusion loss.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
according to a formula containing λ xi+(1-λ)xjObtaining a fused sample image according to the relation; wherein, λ is fusion ratio, xiFor the first training sample image, xjIs the second training sample image.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
according to the formula containing λ cross EntroLoss (y', y)i)+(1-λ)*crossEntroLoss(y',yj) Obtaining the fusion loss according to the relation; wherein y' is the classification result of the car logo, yiIs a first class vector, yjAs a second class vector, crossEntroLoss (y', y)i) For the first classification loss, crossEntroLoss (y', y)j) Is the second classification loss.
In one embodiment, there is also provided a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the following steps when executing the computer program:
acquiring an image to be detected;
detecting an image to be detected by adopting a vehicle detection model, and marking a region of a vehicle if the vehicle is detected; if the vehicle is not detected, the process is ended;
detecting the marked vehicle area by adopting a vehicle logo detection model, and marking the area of the vehicle logo if the vehicle logo is detected; if the car logo is not detected, ending the process;
and identifying the marked car logo area by adopting a car logo classification model to obtain a car logo identification result.
The implementation principle and technical effect of the computer device provided in this embodiment are similar to those of the method embodiments described above, and are not described herein again.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; the training sample image comprises a category label;
inputting the fused sample image into a car logo classification model for classification processing to obtain a car logo classification result;
respectively calculating a first classification loss and a second classification loss of the vehicle logo classification result and the class labels of any two training sample images, and fusing the first classification loss and the second classification loss according to a fusion ratio to obtain a fusion loss;
and adjusting model parameters in the car logo classification model according to the fusion loss, and performing circular training until the car logo classification model converges.
The implementation principle and technical effect of the computer-readable storage medium provided by this embodiment are similar to those of the above-described method embodiment, and are not described herein again.
In one embodiment, the computer program when executed by the processor further performs the steps of:
multiplying a first training sample image in any two training sample images by the fusion proportion to obtain a first result;
multiplying a second training sample image in any two training sample images by the deformation form of the fusion proportion to obtain a second result;
and adding the first result and the second result to obtain a fused sample image.
In one embodiment, the car logo classification result is a probability vector of the fused sample image belonging to different car logo classes; the computer program when executed by the processor further realizes the steps of:
respectively carrying out one-hot transformation on the class labels of any two training sample images to obtain a first class vector and a second class vector;
and calculating a first classification loss of the vehicle logo classification result and the first class vector and a second classification loss of the vehicle logo classification result and the second class vector.
In one embodiment, the computer program when executed by the processor further performs the steps of:
multiplying the first classification loss by the fusion proportion to obtain a third result; multiplying the second classification loss by the deformation form of the fusion proportion to obtain a fourth result;
and adding the third result and the fourth result to obtain fusion loss.
In one embodiment, the computer program when executed by the processor further performs the steps of:
according to a formula containing λ xi+(1-λ)xjObtaining a fused sample image according to the relation; wherein, λ is fusion ratio, xiFor the first training sample image, xjIs the second training sample image.
In one embodiment, the computer program when executed by the processor further performs the steps of:
according to the formula containing λ cross EntroLoss (y', y)i)+(1-λ)*crossEntroLoss(y',yj) Obtaining the fusion loss according to the relation; wherein y' is the classification result of the car logo, yiIs a first class vector, yjAs a second class vector, crossEntroLoss (y', y)i) For the first classification loss, crossEntroLoss (y', y)j) Is the second classification loss.
In one embodiment, there is also provided a computer readable storage medium having a computer program stored thereon, the computer program when executed by a processor implementing the steps of:
acquiring an image to be detected;
detecting an image to be detected by adopting a vehicle detection model, and marking a region of a vehicle if the vehicle is detected; if the vehicle is not detected, the process is ended;
detecting the marked vehicle area by adopting a vehicle logo detection model, and marking the area of the vehicle logo if the vehicle logo is detected; if the car logo is not detected, ending the process;
and identifying the marked car logo area by adopting a car logo classification model to obtain a car logo identification result.
The implementation principle and technical effect of the computer-readable storage medium provided by this embodiment are similar to those of the above-described method embodiment, and are not described herein again.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A training method of a car logo classification model is characterized by comprising the following steps:
acquiring a plurality of training sample images, and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; the training sample image comprises a class label;
inputting the fused sample image into a car logo classification model for classification processing to obtain a car logo classification result;
respectively calculating a first classification loss and a second classification loss of the car logo classification result and the class labels of any two training sample images, and fusing the first classification loss and the second classification loss according to the fusion proportion to obtain a fusion loss;
and adjusting model parameters in the car logo classification model according to the fusion loss, and performing circular training until the car logo classification model converges.
2. The method according to claim 1, wherein the fusing any two training sample images in the plurality of training sample images according to a preset fusion ratio to obtain a fused sample image comprises:
multiplying a first training sample image in the any two training sample images by the fusion proportion to obtain a first result;
multiplying a second training sample image in the any two training sample images by the deformation form of the fusion proportion to obtain a second result;
and adding the first result and the second result to obtain the fusion sample image.
3. The method according to claim 1 or 2, wherein the car logo classification result is a probability vector that the fused sample image belongs to different car logo classes; the calculating a first classification loss and a second classification loss of the car logo classification result and the class labels of any two training sample images respectively comprises:
respectively carrying out one-hot transformation on the class labels of any two training sample images to obtain a first class vector and a second class vector;
and calculating the first classification loss of the vehicle logo classification result and the first class vector and the second classification loss of the vehicle logo classification result and the second class vector.
4. The method of claim 3, wherein said fusing the first classification loss and the second classification loss according to the fusion ratio to obtain a fusion loss comprises:
multiplying the first classification loss by the fusion proportion to obtain a third result; multiplying the second classification loss by the deformation form of the fusion proportion to obtain a fourth result;
and adding the third result and the fourth result to obtain the fusion loss.
5. The method according to claim 2, wherein the fusing any two training sample images in the plurality of training sample images according to a preset fusion ratio to obtain a fused sample image comprises:
according to a formula containing λ xi+(1-λ)xjObtaining the fused sample image according to the relational expression; wherein λ is the fusion ratio, xiFor the first training sample image, the xjIs the second training sample image.
6. The method of claim 4, wherein said fusing the first classification loss and the second classification loss according to the fusion ratio to obtain a fusion loss comprises:
according to the formula containing λ cross EntroLoss (y', y)i)+(1-λ)*crossEntroLoss(y',yj) Obtaining the fusion loss; wherein y' is the car logo classification result, and yiIs the first class vector, the yjIs the second class vector, the crossEntroLoss (y', y)i) For the first classification loss, the crossEntroLoss (y', y)j) Is the second classification loss.
7. A car logo identification method is characterized by comprising the following steps:
acquiring an image to be detected;
detecting the image to be detected by adopting a vehicle detection model, and marking the area of the vehicle if the vehicle is detected; if the vehicle is not detected, the process is ended;
detecting a marked vehicle area by adopting a vehicle logo detection model, and marking the area of the vehicle logo if the vehicle logo is detected; if the car logo is not detected, ending the process;
adopting a vehicle logo classification model to identify the marked vehicle logo area to obtain a vehicle logo identification result; the training method of the car logo classification model is the method of any one of claims 1-6.
8. The utility model provides a trainer of car logo classification model which characterized in that includes:
the device comprises a first acquisition module, a second acquisition module and a fusion module, wherein the first acquisition module is used for acquiring a plurality of training sample images and fusing any two training sample images in the plurality of training sample images according to a preset fusion proportion to obtain a fusion sample image; the training sample image comprises a class label;
the classification module is used for inputting the fusion sample image into a car logo classification model for classification processing to obtain a car logo classification result;
the calculation module is used for calculating a first classification loss and a second classification loss of the car logo classification result and the class labels of any two training sample images respectively, and fusing the first classification loss and the second classification loss according to the fusion proportion to obtain a fusion loss;
and the training module is used for adjusting model parameters in the car logo classification model according to the fusion loss so as to carry out circular training until the car logo classification model converges.
9. A emblem recognition device, comprising:
the second acquisition module is used for acquiring an image to be detected;
the first detection module is used for detecting the image to be detected by adopting a vehicle detection model, and marking the area of the vehicle if the vehicle is detected; if the vehicle is not detected, the process is ended;
the second detection module is used for detecting the marked vehicle area by adopting the vehicle logo detection model, and marking the area of the vehicle logo if the vehicle logo is detected; if the car logo is not detected, ending the process;
and the identification module is used for identifying the marked car logo area by adopting the car logo classification model to obtain a car logo identification result.
10. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method according to any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910875817.6A CN110633751A (en) | 2019-09-17 | 2019-09-17 | Training method of car logo classification model, car logo identification method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910875817.6A CN110633751A (en) | 2019-09-17 | 2019-09-17 | Training method of car logo classification model, car logo identification method, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110633751A true CN110633751A (en) | 2019-12-31 |
Family
ID=68972682
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910875817.6A Pending CN110633751A (en) | 2019-09-17 | 2019-09-17 | Training method of car logo classification model, car logo identification method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110633751A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111369063A (en) * | 2020-03-06 | 2020-07-03 | 北京易真学思教育科技有限公司 | Test paper model training method, test paper combining method and related device |
| CN111444878A (en) * | 2020-04-09 | 2020-07-24 | Oppo广东移动通信有限公司 | Video classification method and device and computer readable storage medium |
| CN112232349A (en) * | 2020-09-23 | 2021-01-15 | 成都佳华物链云科技有限公司 | Model training method, image segmentation method and device |
| CN112418303A (en) * | 2020-11-20 | 2021-02-26 | 浙江大华技术股份有限公司 | Training method and device for recognizing state model and computer equipment |
| CN112837541A (en) * | 2020-12-31 | 2021-05-25 | 遵义师范学院 | Intelligent traffic vehicle flow management method based on improved SSD |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103077407A (en) * | 2013-01-21 | 2013-05-01 | 信帧电子技术（北京）有限公司 | Car logo positioning and recognition method and car logo positioning and recognition system |
| CN103295021A (en) * | 2012-02-24 | 2013-09-11 | 北京明日时尚信息技术有限公司 | Method and system for detecting and recognizing feature of vehicle in static image |
| US20160300337A1 (en) * | 2015-04-08 | 2016-10-13 | Tatung University | Image fusion method and image processing apparatus |
| CN109919888A (en) * | 2019-02-26 | 2019-06-21 | 腾讯科技（深圳）有限公司 | A kind of method of image co-registration, the method for model training and relevant apparatus |
| CN109961051A (en) * | 2019-03-28 | 2019-07-02 | 湖北工业大学 | A kind of pedestrian's recognition methods again extracted based on cluster and blocking characteristic |
| CN110135371A (en) * | 2019-05-20 | 2019-08-16 | 华南农业大学 | A kind of Citrus Huanglongbing pathogen recognition methods and device based on Mixup algorithm |
-
2019
- 2019-09-17 CN CN201910875817.6A patent/CN110633751A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103295021A (en) * | 2012-02-24 | 2013-09-11 | 北京明日时尚信息技术有限公司 | Method and system for detecting and recognizing feature of vehicle in static image |
| CN103077407A (en) * | 2013-01-21 | 2013-05-01 | 信帧电子技术（北京）有限公司 | Car logo positioning and recognition method and car logo positioning and recognition system |
| US20160300337A1 (en) * | 2015-04-08 | 2016-10-13 | Tatung University | Image fusion method and image processing apparatus |
| CN109919888A (en) * | 2019-02-26 | 2019-06-21 | 腾讯科技（深圳）有限公司 | A kind of method of image co-registration, the method for model training and relevant apparatus |
| CN109961051A (en) * | 2019-03-28 | 2019-07-02 | 湖北工业大学 | A kind of pedestrian's recognition methods again extracted based on cluster and blocking characteristic |
| CN110135371A (en) * | 2019-05-20 | 2019-08-16 | 华南农业大学 | A kind of Citrus Huanglongbing pathogen recognition methods and device based on Mixup algorithm |
Non-Patent Citations (2)
| Title |
|---|
| HONGYI ZHANG ET AL.: ""mixup: BEYOND EMPIRICAL RISK MINIMIZATION"", 《ARXIV》 * |
| HONGYI-ZHANG: ""mixup / cifar / utils.py"", 《MIXUP/UTILS.PY AT MASTER · HONGYI-ZHANG/MIXUP · GITHUB》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111369063A (en) * | 2020-03-06 | 2020-07-03 | 北京易真学思教育科技有限公司 | Test paper model training method, test paper combining method and related device |
| CN111369063B (en) * | 2020-03-06 | 2022-08-09 | 北京易真学思教育科技有限公司 | Test paper model training method, test paper combining method and related device |
| CN111444878A (en) * | 2020-04-09 | 2020-07-24 | Oppo广东移动通信有限公司 | Video classification method and device and computer readable storage medium |
| CN112232349A (en) * | 2020-09-23 | 2021-01-15 | 成都佳华物链云科技有限公司 | Model training method, image segmentation method and device |
| CN112232349B (en) * | 2020-09-23 | 2023-11-03 | 成都佳华物链云科技有限公司 | Model training method, image segmentation method and device |
| CN112418303A (en) * | 2020-11-20 | 2021-02-26 | 浙江大华技术股份有限公司 | Training method and device for recognizing state model and computer equipment |
| CN112837541A (en) * | 2020-12-31 | 2021-05-25 | 遵义师范学院 | Intelligent traffic vehicle flow management method based on improved SSD |
| CN112837541B (en) * | 2020-12-31 | 2022-04-29 | 遵义师范学院 | Intelligent traffic vehicle flow management method based on improved SSD |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110633751A (en) | Training method of car logo classification model, car logo identification method, device and equipment | |
| WO2021135499A1 (en) | Damage detection model training and vehicle damage detection methods, device, apparatus, and medium | |
| TWI742382B (en) | Neural network system for vehicle parts recognition executed by computer, method for vehicle part recognition through neural network system, device and computing equipment for vehicle part recognition | |
| US20210295162A1 (en) | Neural network model training method and apparatus, computer device, and storage medium | |
| CN111950329A (en) | Target detection and model training method and device, computer equipment and storage medium | |
| CN110674712A (en) | Interactive behavior recognition method and device, computer equipment and storage medium | |
| CN110807491A (en) | License plate image definition model training method, definition detection method and device | |
| CN110533086B (en) | Semi-automatic image data labeling method | |
| CN110942012A (en) | Image feature extraction method, pedestrian re-identification method, device and computer equipment | |
| CN111832581B (en) | Lung feature recognition method and device, computer equipment and storage medium | |
| CN113128478B (en) | Model training method, pedestrian analysis method, device, equipment and storage medium | |
| CN110046577B (en) | Pedestrian attribute prediction method, device, computer equipment and storage medium | |
| CN111144398A (en) | Target detection method, target detection device, computer equipment and storage medium | |
| CN110796634A (en) | Dam state detection method and device, computer equipment and readable storage medium | |
| CN110569856A (en) | sample labeling method and device, and damage category identification method and device | |
| CN111914842A (en) | License plate information identification method and device, computer equipment and storage medium | |
| CN113807399A (en) | Neural network training method, neural network detection method and neural network detection device | |
| CN111047088A (en) | Prediction image acquisition method and device, computer equipment and storage medium | |
| CN111178358A (en) | Text recognition method and device, computer equipment and storage medium | |
| CN110717449A (en) | Vehicle annual inspection personnel behavior detection method and device and computer equipment | |
| CN112001399A (en) | Image scene classification method and device based on local feature saliency | |
| CN116363357A (en) | Semi-supervised semantic segmentation method and device based on MIM and contrast learning | |
| CN111898566A (en) | Attitude estimation method, attitude estimation device, electronic equipment and storage medium | |
| CN113496148A (en) | Multi-source data fusion method and system | |
| CN111178162B (en) | Image recognition method, device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191231 |